Santiago Berrezueta wins "Best Presentation" award at the 2024 IE Conference

Tuesday, 25 June 2024 • Ignacio Garcia

The International Conference on Intelligent Environments (IE) is held annually at different locations around the world, and focuses on research in the field of interactive spaces made possible by the use of different technologies. Artificial intelligence, communication technologies, and sensing interfaces, along with more contributions from multiple disciplines, allow for the design of intelligent environments that can enrich users’ activities and interactions.

The 20th edition of the conference was held during June in Ljubljana, Slovenia, and awarded distinctions to the best papers and presentations. Santiago Berrezueta was awarded the "Best Presentation Award" for his talk based on the paper “Exploring the Efficacy of Robotic Assistants with ChatGPT and Claude in Enhancing ADHD Therapy: Innovating Treatment Paradigms”, co-authored by himself, Mohanad Kandil, María-Luisa Martín-Ruiz, Iván Pau-de-la-Cruz, and Stephan Krusche.

Award for the best presentation

The paper explores the possibility of integrating large language models into robotic assistants used for ADHD therapy. Attention Deficit Hyperactivity Disorder (ADHD) is a neurodevelopmental condition that can significantly impact an individual’s daily functioning. Occupational therapy plays a crucial role in managing ADHD, as it helps individuals improve their ability to fully participate in school, home, and social activities. Considering the fact that there are recent studies that highlight the potential of integrating Large Language Models into psychological treatments to improve the treatments’ efficiency, the paper analyzes the potential of LLMS to be integrated into occupational therapy for ADHD.

Two LLMs, ChatGPT-4 Turbo and Claude-3 Opus, were integrated into a robotic assistant to explore how well each model performs in robot-assisted interactions. Their performance, adaptability, and responsiveness was analyzed in this context, and compared to a clinically validated customized model. The results demonstrated that both models offered coherent and safe interactions, and could be selected to be integrated into a robotic assistant based on the specific demands of the ADHD therapy.

Information on the paper, as well as a freely accessible preprint, can be found below: